Re: mapping to WSDL (e.g. wsdl:endpoint)?

On Mon, May 09, 2005 at 11:36:31PM +0900, Yoshio FUKUSHIGE wrote:
> 
> Hi,
> 
> <!--
> (Reading your message on cancelled telecon)
> So you don't seem to be here in Japan now.
> It's a pity, but I think we can still discuss by email.

Bijan Parsia is in Japan for two weeks at WWW05 and visiting one of
our lab partners (NTT or someone, not sure), and he's also responsible
for the WSDL->RDF mapping, so he'd be a good person to talk with
directly.

> In the SADDLE example in the protocol doc, it says:
> 
> _:sericeA
>    wsdl:endpoint <http://a.example/sparql/>;
>    wsdl:interface sl:SparqlQuery.
> 
> and like that.
> 
> On the other hand, in WSDL2.0, an "endpoint" is expressed as an element 
> that has
> several required properties (name, binding) and optional properties and 
> elements(address, features, properties, documentation)

Yes.

> I wonder if we are going to describe a SPARQL service
> using RDF (with SADDLE vocabulary) while maintaining the same abstract 
> model as WSDL.

I think that what we need in SADDLE is a way to say, for a Service S,
that for some pair of URIs U1 and U2, U1 should be dereferenced to
retrieve a representation of the WSDL description of S, and U2 should
be dereferenced to retrieve a representation of the SADDLE description
of S. WSDL contains the service details -- what interfaces and
operations it contains, their inputs and outputs, and the concrete
details of *where* and *how* to invoke those operatiosn -- and SADDLE
contains the details specific to the kind of service, SPARQL query, S
is.

If this were 2007, we could put all of the SADDLE information into the
extensibility framework of WSDL 2.0, thus conceivably collapsing U1
and U2 into a single URI. But that stuff is radically unbaked, IMO,
and too far away for us to wait on. (But I'm not an expert
hereabouts...)

> If so, I don't know how binding information is described with such 
> wsdl:endpoint predicate.

It isn't. Bindings are described by, in my example above, U1, the URI
identifying the WSDL describing S.

> Please guide me to right direction if I see wrong direction.

No, you seem to understand the issues as well as anyone. I'm not sure
what to say to help you fix the problem just yet.

My preferred design is that a SADDLE description contain a predicate
pointing to the WSDL for that service. And vice versa; that is, the
WSDL I'm proposing will contain an interface and operation, which
services can implement, in order to retrieve the SADDLE
(SparqlService.getSADDLE).

So if you have the SADDLE (URI or representation), you (or a client)
can find the WSDL and invoke some operations; conversely, if you have
the WSDL description, and thus can invoke operations, ideally you can
retrieve the SADDLE description, thus learning more domain-specific
detail about the service at hand.

> Maybe I miss the description of the wsdl: vocabulary, in that case,
> please give me a pointer.
> ...7.15 Mapping to RDF and Semantic Web of the WSDL 2.0 Primer is still 
> empty : (

Yes, I know. I'm going to CC: Bijan Parsia on this message, so that
perhaps you and he can talk about the WSDL->RDF mapping face to face
in Japan.

Best,
Kendall Clark

Received on Monday, 9 May 2005 15:15:25 UTC